Er. I think you have made a mistake in your benchmarking. You swapped the labels for the code used for the "sub" and "builtin".

I fixed the benchmark and under perl 5.6.1 get the results one would expect:

use Benchmark 'cmpthese';

sub mychr { return pack("C", shift); }

@array = (0 .. 255) x 100;


cmpthese(-2, {
  'builtin'     => 'foreach (@array) { my $chr = chr($_) }',
  'sub'         => 'foreach (@array) { my $chr = &mychr($_) }',
  'pack'        => 'foreach (@array) { my $chr = pack("C", $_) }',
});
__END__
Benchmark: running builtin, pack, sub, each for at least 2 CPU seconds
+...
   builtin:  2 wallclock secs ( 2.06 usr +  0.00 sys =  2.06 CPU) @ 36
+.84/s (n=76)
      pack:  2 wallclock secs ( 2.11 usr +  0.00 sys =  2.11 CPU) @ 27
+.03/s (n=57)
       sub:  2 wallclock secs ( 2.06 usr +  0.00 sys =  2.06 CPU) @  8
+.24/s (n=17)
          Rate     sub    pack builtin
sub     8.24/s      --    -70%    -78%
pack    27.0/s    228%      --    -27%
builtin 36.8/s    347%     36%      --
[download]

Cheers,

Yves / DeMerphq
---
Writing a good benchmark isnt as easy as it might look.